CN104023233B - Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) - Google Patents
Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) Download PDFInfo
- Publication number
- CN104023233B CN104023233B CN201410288532.XA CN201410288532A CN104023233B CN 104023233 B CN104023233 B CN 104023233B CN 201410288532 A CN201410288532 A CN 201410288532A CN 104023233 B CN104023233 B CN 104023233B
- Authority
- CN
- China
- Prior art keywords
- current
- thr
- hevc
- skip
- depth
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a fast inter-frame prediction method of HEVC (High Efficiency Video Coding). The method comprises the steps that two sets of robust threshold values are obtained through statistics to meet the different depths of coding units (CUs), and different quantization parameters (QP); when the rate distortion cost of the SKIP mode of a current CU is less than a given threshold value, the optimal prediction mode of the CU is a SKIP mode; when the rate distortion cost of the optimal prediction mode of the current CU is less than another given threshold value, the CU stops further segmentation. According to the fast inter-frame prediction method disclosed by the invention, the unnecessary deep calculation of the coding units can be precisely terminated, the unnecessary prediction mode calculation can be omitted, and the inter-frame coding complexity of HEVC can be greatly lowered, which contributes to the real-time application of an HEVC encoder.
Description
Technical field
The present invention relates to the video information process field in digital video communications field, more particularly to a kind of HEVC fast frames
Between Forecasting Methodology.
Background technology
With the continuous progress of science and technology, people are to vision and the requirement more and more higher of acoustical quality, HD video and superelevation
Clear video starts by common concern.H.264/AVC, the second generation video encoding standard founded before for 10 years can not meet people
For the requirement of practical application, the serious hope of industrial quarters and academia to video encoding standard of new generation is more and more stronger.At this
Under situation, VCEG and the big normal structures of MPEG two proceed by cooperative development, and referred to as JCT-VC has been set up in January, 2010
The united organization of (Joint Collaborative Team on Video Coding), starts unified next generation's video of formulating and compiles
Code standard is simultaneously named as High Efficiency Video Coding (HEVC, high efficiency Video coding), and in January, 2013
No. 26 formally turn into International video coding standard.
Although HEVC uses block-based coding framework, HEVC to do largely as conventional video encoding standard
Technological innovation.Three coding unit concepts are proposed first：CU (Coding Unit, coding unit), PU (Prediction
Unit, predicting unit), TU (Transform Unit, converter unit).For CU innovatively using the block based on quaternary tree
Size recursive subdivision structure, full-size is up to 64 × 64.As shown in figure 1, for inter prediction, the CU of each depth level is (i.e.
CU0、CU1、CU2、CU3) there are SKIP (skip mode, the pattern is without residual error), 2N × 2N (not Fractionation regimen), 2N × N (length
2 times of width of Fractionation regimen), the inter-frame forecast mode such as N × 2N (width is the Fractionation regimen of 2 times of length)；If CU
Depth be 3, then the inter-frame forecast mode of the CU also include AMP (asymmetric segmentation) pattern：2N × nU is (more than the pattern
Fang Changdu is 4 times of width, lower width is length 3/4 divides), (pattern is length with upper width to 2N × nD
3/4th, lower section length is 4 times of width and divides), (pattern is 4 times of width with left length to nL × 2N, right side width is
The 3/4 of length divides), (pattern is the 3/4 of length with left side width to nR × 2N, right length is 4 times of width and draws
Point).It is referred to as RQT (Residual Quad-tree Transform, residual error quaternary tree) segmenting structure using a kind of for TU, its
The big I of transform block carries out the adjustment of self adaptation according to the characteristic of residual error.Above-mentioned listed technology is the numerous of HEVC introductions
Part the most prominent in technology.These innovative technologies introduced cause that HEVC is ensureing same video quality with H.264 comparing
While, code check can reduce 50% or so.It can be seen that, with the continuous demand to following high definition and ultra high-definition video, Yi Jiyi
Dynamic equipment is greatly developed, and on the premise of available bandwidth is limited, HEVC has shown huge market application potential.But taking
While obtaining high-gain, HEVC is also faced with a huge problem, and with H.264 comparing, the encoder complexity of HEVC is at least
Its 4 times, it is unfavorable for real-time application.And the model selection in HEVC accounts for more than the 80% of the whole scramble time, it is seen that research
A kind of efficient fast mode decision algorithm is particularly important for reducing the encoder complexity of HEVC.
For the high-speed decision of interframe CU and PU, numerous motions are had been presented at present, such as the subtree based on SKIP patterns
Pruning Algorithm in advance, the method first checks for whether current CU is SKIP patterns, if then terminating its cutting procedure, the party in advance
Method is easy and effective, referred to as ECU；In addition a kind of CFM is also proposed, that is, checks present intensity block and corresponding two chrominance blocks
Cbf (coded block flag) whether all 0, if so, then terminating the remaining PU mode treatments of current CU depth；It is same to go back
A kind of method for detecting SKIP patterns in advance is proposed, estimation is carried out to 2N × 2N patterns first, its DMV for checking afterwards
(motion vector difference) and cbf whether all 0, if then current PU optimization models are SKIP patterns, should
Method is referred to as ESD.Current these three methods, are adopted by HEVC standard, but the encoder complexity of HEVC is still quite huge.This
Invention provides one kind and is applied to the simple and highly efficient quick CU of interframe and PU decision-making techniques with the angle different from this three, this
Inventive method helps to realize the real-time application of HEVC encoders.
It is briefly described as follows the inter predication process of HEVC.As shown in figure 1, coding unit of the division of interframe CU from maximum
I.e. 64 × 64 CU starts, and now depth is 0.Coding is predicted to it first, its RDcost (Rate- is obtained
Distortion cost, rate distortion costs), it is then divided into 4 sub- CU, every size of sub- CU is 32 × 32, and depth is
1, coding equally is predicted to this 4 sub- CU respectively, respectively obtain respective RDcost.So recursively divide；When
When the depth of CU is 3, i.e., when size is 8 × 8, terminate the segmentation of current CU.Then the pattern split is selected, i.e., successively
Whether the RDcost sums for comparing 48 × 8 are less than the RDcost of its corresponding 16 × 16 CU, if it is, selection 8 × 8
Fractionation regimen, otherwise selects 16 × 16 Fractionation regimen.So compare down, until the depth of current CU is 0.Additionally, right
In each depth level, it is SKIP that each CU needs in PU patterns simultaneously, and RDO (Rate are carried out in 2N × 2N, N × 2N and 2N × N
Distortion Optimization, rate-distortion optimization) calculate, if the depth of current CU is not 3, then must be to institute yet
Some AMP patterns：2N × nU, 2N × nD, nL × 2N and nR × 2N carry out RDO calculating, obtain all predictive modes for being calculated
Each rate distortion costs.Then, for the CU of each depth level, a RDcost minimum is chosen in all PU patterns of the CU
Pattern as the CU optimal prediction modes.
From the above, it can be seen that inter prediction needs CU and each depth exhaustively to 4 depth levels (scope is 0-3)
The all PU patterns spent on the CU of level carry out RDO calculating, can just determine optimal Fractionation regimen, it is seen that complexity is quite huge
Greatly.Therefore, a kind of effective method is invented to be particularly important to reduce the computation complexity of HEVC inter predictions.
The content of the invention
A kind of deficiency it is an object of the invention to overcome prior art, there is provided quick inter-frame prediction methods of HEVC, is based on
The rate distortion costs of coded block and movable information vector, can accurately terminate the unnecessary depth calculation of coding unit and jump
Cross unnecessary predictive mode to calculate, the interframe encode complexity of HEVC can be greatly reduced, help to realize that HEVC is encoded
The real-time application of device.
The technical solution adopted for the present invention to solve the technical problems is：A kind of quick inter-frame prediction methods of HEVC, first
Coding unit is designated as CU, quantization parameter is designated as QP；Then CU that depth level is 0,1 and 2 is counted respectively under different Q P
The threshold condition that the rate distortion costs needs of the optimal prediction modes of the CU meet when not split further, obtains right respectively
Answer three groups of threshold values of 0,1,2 depth levels as follows：
Thr64=2270 × e0.8907×QP (1)
Thr32=722.2 × e0.1096×QP (2)
Thr16=228.5 × e0.1136×QP (3)
Wherein, e is math constant, is the truth of a matter of natural logrithm；Afterwards, the CU that depth level is 0,1,2 and 3 is counted respectively
The rate distortion costs of its SKIP pattern need the threshold for meeting when the rate-distortion optimization that SKIP patterns are only carried out under different Q P is calculated
Value condition, four groups of threshold values that 0,1,2,3 depth levels are corresponded to respectively are as follows：
thr64×64=102.6 × e0.1636×QP (4)
thr32×32=9.685 × e0.1888×QP (5)
thr16×16=1.945 × e0.2174×QP (6)
thr8×8=1.218 × e0.212×QP (7)
Wherein, e is math constant, is the truth of a matter of natural logrithm；0,1,2,3 depth levels of CU correspond to 64 × 64 respectively again,
32 × 32,16 × 16,8 × 8 CU sizes；Then the quick inter-frame prediction method comprises the following steps：
1) a maximum sized CU is read；Go to step 2)；
2) the SKIP patterns to current CU carry out rate-distortion optimization calculating, obtain its rate distortion costs, are designated as SKIP_
Cost, then judges the depth of current CU；If 0, go to step 3)；If 1, go to step 4)；If 2, go to step 5)；
Otherwise go to step 6)；
3) thr is calculated according to its QP value set in advance by formula (4)64×64, and whether judge current SKIP_cost
Less than the thr64×64；If so, going to step 11)；Otherwise go to step 7)；
4) thr is calculated according to its QP value set in advance by formula (5)32×32, and whether judge current SKIP_cost
Less than the thr32×32；If so, going to step 11)；Otherwise go to step 7)；
5) thr is calculated according to its QP value set in advance by formula (6)16×16, and whether judge current SKIP_cost
Less than the thr16×16；If so, going to step 11)；Otherwise go to step 7)；
6) thr is calculated according to its QP value set in advance by formula (7)8×8, and judge whether current SKIP_cost is small
In the thr8×8；If so, going to step 11)；Otherwise go to step 7)；
7) interframe 2N × 2N (N=32,16, the 8 or 4) predictive mode to current CU carries out rate-distortion optimization calculating, obtains
Its rate distortion costs；Then the fortune of current CU, the CU of current CU adjacent left-hands, current CU CU these three CU adjacent and above is judged
Whether dynamic vector is all 0；If so, going to step 11)；Otherwise, step 8 is gone to)；
8) respectively to interframe N × 2N (N=32,16,8 or 4) predictive modes and interframe 2N × N of current CU (N=32,16,
8 or 4) predictive mode carries out rate-distortion optimization calculating, obtain their rate distortion costs；Then judge that current CU is pre- in 2N × 2N
Whether the motion vector under survey pattern is 0；If so, going to step 10), otherwise go to step 9)；
If 9) depth of current CU is not 3, rate distortion is carried out to all interframe AMP predictive modes of current CU respectively excellent
Change and calculate, obtain each rate distortion costs of all interframe AMP predictive modes of current CU, then go to step 10)；If current CU's
Depth is 3, then pass directly to step 10)；
10) all intra prediction modes respectively to current CU carry out rate-distortion optimization calculating, obtain all of current CU
Each rate distortion costs of intra prediction mode；Then go to step 11)；
11) each rate distortion costs of the relatively more current calculated all predictive modes of CU, by wherein minimum rate distortion generation
Valency is designated as RDcost_best, and the predictive mode corresponding to the minimum rate distortion costs is recorded as the optimum prediction of current CU
Pattern；Then the depth of current CU is judged；If 0, go to step 12)；If 1, go to step 13)；If 2, go to step
14)；Next maximum sized CU is otherwise read, and goes to step 2)；
12) Thr is calculated according to its QP value set in advance by formula (1)64, then judge that current RDcost_best is
It is no less than the Thr64；If so, going to step 15)；Otherwise go to step 16)；
13) Thr is calculated according to its QP value set in advance by formula (2)32, then judge that current RDcost_best is
It is no less than the Thr32；If so, going to step 15)；Otherwise go to step 16)；
14) Thr is calculated according to its QP value set in advance by formula (3)16, then judge that current RDcost_best is
It is no less than the Thr16；If so, going to step 15)；Otherwise go to step 16)；
15) terminate the further segmentation of current CU, extract next maximum sized CU, then go to step 2)；
16) current CU is divided into 4 sub- CU of size identical, step 2 is gone to respectively to every sub- CU) carry out accordingly
Treatment.
In one embodiment：The span of the quantization parameter is 0 to 51.
In one embodiment：The size of the maximum sized CU is 64 × 64.
From the above-mentioned description of this invention, compared with prior art, the present invention has the advantages that：
1. quick inter-frame prediction methods of a kind of HEVC of the invention, count two groups of healthy and strong threshold values (i.e. formula (1) first
To (3) and formula (4) to (7)) coding unit (CU) of different depth level and different quantization parameters (QP) can be met；When
When the rate distortion costs of the SKIP patterns of current CU are less than given threshold value, the optimal prediction modes of the CU are SKIP patterns；When work as
When the rate distortion costs of the optimal prediction modes of preceding CU are less than another given threshold value, the CU stops performing further segmentation.This
The method of invention is easy and effective, can accurately terminate the unnecessary division of current coded unit and skip current unnecessary pre-
Mode computation is surveyed, the interframe encode complexity of HEVC can be greatly reduced, and can flexibly control selected threshold value, be applicable it
In different application scenarios, while the method for the present invention is very beneficial for the realization of hardware and software.
2. the present invention can't change the original code flow structures of HEVC, completely compatible with HEVC standard.
Brief description of the drawings
Fig. 1 is the recurrence partition process schematic diagram of the interframe encode unit of HEVC.
Specific embodiment
Embodiment,
The invention provides a kind of quick inter-frame prediction methods of HEVC, coding unit is designated as CU first, by quantization parameter
It is designated as QP；Then the optimal of the CU when CU that depth level is 0,1 and 2 is not split further under different Q P is counted respectively
The rate distortion costs of predictive mode need the threshold condition for meeting, and three groups of threshold values that 0,1,2 depth levels are corresponded to respectively are as follows：
Thr64=2270 × e0.8907×QP (1)
Thr32=722.2 × e0.1096×QP (2)
Thr16=228.5 × e0.1136×QP (3)
Wherein, e represents exponential function；Afterwards, the CU that depth level is 0,1,2 and 3 is counted respectively only to enter under different Q P
The rate distortion costs of its SKIP pattern need the threshold condition for meeting when the rate-distortion optimization of row SKIP patterns is calculated, and are distinguished
Four groups of threshold values of 0,1,2,3 depth levels of correspondence are as follows：
thr64×64=102.6 × e0.1636×QP (4)
thr32×32=9.685 × e0.1888×QP (5)
thr16×16=1.945 × e0.2174×QP (6)
thr8×8=1.218 × e0.212×QP (7)
Wherein, e represents exponential function；0,1,2,3 depth levels of CU correspond to 64 × 64,32 × 32,16 × 16,8 respectively again
× 8 CU sizes；Then the quick inter-frame prediction method comprises the following steps：
1) a maximum sized CU (i.e. code tree unit CTU) is read；Go to step 2)；
2) the SKIP patterns to current CU carry out rate-distortion optimization calculating, obtain its rate distortion costs, are designated as SKIP_
Cost, then judges the depth of current CU；If 0, go to step 3)；If 1, go to step 4)；If 2, go to step 5)；
Otherwise go to step 6)；
3) thr is calculated according to its QP value set in advance by formula (4)64×64, and whether judge current SKIP_cost
Less than the thr64×64；If so, going to step 11)；Otherwise go to step 7)；
4) thr is calculated according to its QP value set in advance by formula (5)32×32, and whether judge current SKIP_cost
Less than the thr32×32；If so, going to step 11)；Otherwise go to step 7)；
5) thr is calculated according to its QP value set in advance by formula (6)16×16, and whether judge current SKIP_cost
Less than the thr16×16；If so, going to step 11)；Otherwise go to step 7)；
6) thr is calculated according to its QP value set in advance by formula (7)8×8, and judge whether current SKIP_cost is small
In the thr8×8；If so, going to step 11)；Otherwise go to step 7)；
7) interframe 2N × 2N (as shown in figure 1, N=32,16, the 8 or 4) predictive mode to current CU carries out rate-distortion optimization
Calculate, obtain its rate distortion costs；Then judge current CU, the CU of current CU adjacent left-hands, current CU CU adjacent and above this
Whether three kinds of motion vectors of CU are all 0；If so, going to step 11)；Otherwise, step 8 is gone to)；
8) respectively to interframe N × 2N (as shown in figure 1, N=32,16,8 or 4) predictive modes and interframe 2N × N of current CU
(as shown in figure 1, N=32,16,8 or 4) predictive mode carries out rate-distortion optimization calculating, obtains their rate distortion costs；Then
Judge whether motion vectors of the current CU under 2N × 2N predictive modes is 0；If so, going to step 10), otherwise go to step
9)；
9) if the depth of current CU is not 3, all interframe AMP predictive modes to current CU are (i.e. as shown in Figure 1 respectively
2N × nU, 2N × nD, nL × 2N and nR × 2N this 4 kinds of patterns) carry out rate-distortion optimization calculating, obtain all interframe of current CU
Each rate distortion costs of AMP predictive modes, then go to step 10)；If the depth of current CU is 3, step is passed directly to
10)；
10) all intra prediction modes (i.e. 35 kinds of intra prediction modes of HEVC regulations) respectively to current CU carry out rate
Aberration optimizing is calculated, and obtains each rate distortion costs of all intra prediction modes of current CU；Then go to step 11)；
11) each rate distortion costs of the relatively more current calculated all predictive modes of CU, by wherein minimum rate distortion generation
Valency is designated as RDcost_best, and the predictive mode corresponding to the minimum rate distortion costs is recorded as the optimum prediction of current CU
Pattern；Then the depth of current CU is judged；If 0, go to step 12)；If 1, go to step 13)；If 2, go to step
14)；Next maximum sized CU is otherwise read, and goes to step 2)；
12) Thr is calculated according to its QP value set in advance by formula (1)64, then judge that current RDcost_best is
It is no less than the Thr64；If so, going to step 15)；Otherwise go to step 16)；
13) Thr is calculated according to its QP value set in advance by formula (2)32, then judge that current RDcost_best is
It is no less than the Thr32；If so, going to step 15)；Otherwise go to step 16)；
14) Thr is calculated according to its QP value set in advance by formula (3)16, then judge that current RDcost_best is
It is no less than the Thr16；If so, going to step 15)；Otherwise go to step 16)；
15) terminate the further segmentation of current CU, extract next maximum sized CU, then go to step 2)；
16) current CU is divided into 4 sub- CU of size identical, step 2 is gone to respectively to every sub- CU) carry out accordingly
Treatment.
It is pointed out that the foundation of step of the present invention is, counted in advance current CU, the CU of current CU adjacent left-hands,
The motion vector of current CU CU these three CU adjacent and above and the relation of various predictive modes, and drawn when this 3 motions
Vector it is all 0 when, the predictive mode of 2N × 2N is the probability highest of optimal prediction modes；Meanwhile, current CU has been counted in advance
Motion vector and various predictive modes relation, and drawn when motion vectors of the current CU under 2N × 2N predictive modes is
When 0, AMP patterns are not the probability highest of optimal prediction modes.
In one embodiment：The span of the quantization parameter is 0 to 51.
In one embodiment：The size of the maximum sized CU is 64 × 64.
The effect of the inventive method is weighed using the reference software HM 10.1 of the video encoding standard HEVC offers of a new generation
Really.To HEVC suggestion 5 level of resolution (Class A, B, C, D, E) 18 cycle tests (Traffic,
PeopleOnStreet、ParkScene、BasketballDrive、Cactus、BQTerrace、Kimono1、PartyScene、
RaceHorses、BQMall、BasketballDrill、BasketballPass、BQSquare、RaceHorses、
BlowingBubbles, FourPeople, Johnny, KristenAndSara) all tests.Coder parameters are：LDP
(low-delay-P, low time delay P frame) coding mode, chooses QP=22,27,32,37 respectively, and sequence all encodes 50 frames, remaining
It is default setting.The inventive method is as shown in table 1 with the experimental result that the HEVC primal algorithms that software HM10.1 is used compare.
Wherein, the interpretational criteria that the BDBR in table 1 and BDPSNR is proposed with for VCEG, BDBR is represented two under same objective quality
The code check of the method for kind saves situation, and BDPSNR represents Y-PSNR (the brightness Y peak values letters of the two methods under given equal code check
Make an uproar ratio) difference；△ Time in table 1 are defined as follows：
Wherein TimeHM10.1(QPi) and Timepro(QPi) original method of software HM10.1 and side of the invention are represented respectively
Method with QP value changes scramble time.
Table 1
Be can be seen that from upper table, a kind of quick inter-frame prediction methods of HEVC proposed by the present invention work well, hardly shadow
The distortion performance of encoder is rung, BDBR only increases by 1.011%.It is pointed out that the threshold of formula (1) to (7) of the invention
Value is to meet the threshold value that BDBR 1% or so (as shown in table 1) of increase is drawn；For different application scenarios, the present invention can
The flexible selected threshold value of control, makes it be applied to different application occasion.And, hardware of the present invention is simple with the realization of software, no
Extra computation complexity can be introduced as other method.
Analysis and experimental verification more than, it can be deduced that to draw a conclusion：Theoretical foundation of the present invention is correct, practical application
It is feasible, help to realize the real-time application of HEVC encoders.
Above-described embodiment is only used for further illustrating the quick inter-frame prediction methods of a kind of HEVC of the invention, but the present invention is simultaneously
It is not limited to embodiment, every any simple modification, equivalent change made to above example according to technical spirit of the invention
Change and modify, each fall within the protection domain of technical solution of the present invention.
Claims (3)
1. quick inter-frame prediction methods of a kind of HEVC, it is characterised in that coding unit is designated as CU first, quantization parameter is designated as
QP；Then the optimum prediction of the CU when CU that depth level is 0,1 and 2 is not split further under different Q P is counted respectively
The rate distortion costs of pattern need the threshold condition for meeting, and three groups of threshold values that 0,1,2 depth levels are corresponded to respectively are as follows：
Thr64=2270 × e0.8907×QP (1)
Thr32=722.2 × e0.1096×QP (2)
Thr16=228.5 × e0.1136×QP (3)
Wherein, e is math constant, is the truth of a matter of natural logrithm；Afterwards, CU that depth level is 0,1,2 and 3 is counted respectively not
The rate distortion costs of its SKIP pattern need the threshold value bar for meeting when being calculated with the rate-distortion optimization that SKIP patterns are only carried out under QP
Part, four groups of threshold values that 0,1,2,3 depth levels are corresponded to respectively are as follows：
thr64×64=102.6 × e0.1636×QP (4)
thr32×32=9.685 × e0.1888×QP (5)
thr16×16=1.945 × e0.2174×QP (6)
thr8×8=1.218 × e0.212×QP (7)
Wherein, e is math constant, is the truth of a matter of natural logrithm；0,1,2,3 depth levels of CU correspond to 64 × 64,32 respectively again ×
32nd, 16 × 16,8 × 8 CU sizes；Then the quick inter-frame prediction method comprises the following steps：
1) a maximum sized CU is read；Go to step 2)；
2) the SKIP patterns to current CU carry out rate-distortion optimization calculating, obtain its rate distortion costs, are designated as SKIP_cost, so
The depth of current CU is judged afterwards；If 0, go to step 3)；If 1, go to step 4)；If 2, go to step 5)；Otherwise turn
To step 6)；
3) thr is calculated according to its QP value set in advance by formula (4)64×64, and judge current SKIP_cost whether less than this
thr64×64；If so, going to step 11)；Otherwise go to step 7)；
4) thr is calculated according to its QP value set in advance by formula (5)32×32, and judge current SKIP_cost whether less than this
thr32×32；If so, going to step 11)；Otherwise go to step 7)；
5) thr is calculated according to its QP value set in advance by formula (6)16×16, and judge current SKIP_cost whether less than this
thr16×16；If so, going to step 11)；Otherwise go to step 7)；
6) thr is calculated according to its QP value set in advance by formula (7)8×8, and judge current SKIP_cost whether less than this
thr8×8；If so, going to step 11)；Otherwise go to step 7)；
7) the interframe 2N × 2N predictive modes to current CU carry out rate-distortion optimization calculating, obtain its rate distortion costs；Then sentence
Whether the motion vector of disconnected current CU, the CU of current CU adjacent left-hands, current CU CU these three CU adjacent and above is all 0；
If so, going to step 11)；Otherwise, step 8 is gone to)；
8) the interframe N × 2N predictive modes and interframe 2N × N predictive modes respectively to current CU carry out rate-distortion optimization calculating, obtain
To their rate distortion costs；Then judge whether motion vectors of the current CU under 2N × 2N predictive modes is 0；If so, turning
To step 10), otherwise go to step 9)；
If 9) depth of current CU is not 3, all interframe AMP predictive modes to current CU carry out rate-distortion optimization meter respectively
Calculate, obtain each rate distortion costs of all interframe AMP predictive modes of current CU, then go to step 10)；If the depth of current CU
It is 3, then passes directly to step 10)；
10) all intra prediction modes respectively to current CU carry out rate-distortion optimization calculating, obtain all frame ins of current CU
Each rate distortion costs of predictive mode；Then go to step 11)；
11) each rate distortion costs of the relatively more current calculated all predictive modes of CU, by wherein minimum rate distortion costs note
It is RDcost_best, and the predictive mode corresponding to the minimum rate distortion costs is recorded as the optimal prediction modes of current CU；
Then the depth of current CU is judged；If 0, go to step 12)；If 1, go to step 13)；If 2, go to step 14)；It is no
Next maximum sized CU is then read, and goes to step 2)；
12) Thr is calculated according to its QP value set in advance by formula (1)64, then judge whether current RDcost_best is small
In the Thr64；If so, going to step 15)；Otherwise go to step 16)；
13) Thr is calculated according to its QP value set in advance by formula (2)32, then judge whether current RDcost_best is small
In the Thr32；If so, going to step 15)；Otherwise go to step 16)；
14) Thr is calculated according to its QP value set in advance by formula (3)16, then judge whether current RDcost_best is small
In the Thr16；If so, going to step 15)；Otherwise go to step 16)；
15) terminate the further segmentation of current CU, extract next maximum sized CU, then go to step 2)；
16) current CU is divided into 4 sub- CU of size identical, step 2 is gone to respectively to every sub- CU) located accordingly
Reason.
2. a kind of quick inter-frame prediction methods of HEVC as claimed in claim 1, it is characterised in that the value of the quantization parameter
Scope is 0 to 51.
3. a kind of quick inter-frame prediction methods of HEVC as claimed in claim 1 or 2, it is characterised in that described maximum sized
The size of CU is 64 × 64.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410288532.XA CN104023233B (en) | 2014-06-24 | 2014-06-24 | Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410288532.XA CN104023233B (en) | 2014-06-24 | 2014-06-24 | Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104023233A CN104023233A (en) | 2014-09-03 |
| CN104023233B true CN104023233B (en) | 2017-05-24 |
Family
ID=51439775
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410288532.XA Active CN104023233B (en) | 2014-06-24 | 2014-06-24 | Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104023233B (en) |
Families Citing this family (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104320658B (en) * | 2014-10-20 | 2017-12-26 | 南京邮电大学 | A kind of HEVC fast encoding methods |
| CN104333755B (en) * | 2014-10-27 | 2017-11-14 | 上海交通大学 | The CU based on SKIP/Merge RD Cost of B frames shifts to an earlier date terminating method in HEVC |
| CN104320656B (en) * | 2014-10-30 | 2019-01-11 | 上海交通大学 | Interframe encoding mode fast selecting method in x265 encoder |
| WO2016203981A1 (en) * | 2015-06-16 | 2016-12-22 | シャープ株式会社 | Image decoding device and image encoding device |
| CN105407357B (en) * | 2015-12-04 | 2019-07-05 | 上海交通大学 | SKIP mode quick selecting method based on Neyman-Pearson rule |
| CN105578197B (en) * | 2015-12-24 | 2019-04-30 | 福州瑞芯微电子股份有限公司 | A kind of realization inter-prediction master control system |
| CN105578195B (en) * | 2015-12-24 | 2019-03-12 | 福州瑞芯微电子股份有限公司 | A kind of H.264 inter-frame prediction system |
| CN105721865A (en) * | 2016-02-01 | 2016-06-29 | 同济大学 | Fast decision algorithm for dividing HEVC inter-frame coding unit |
| US10511859B2 (en) * | 2016-02-04 | 2019-12-17 | Mediatek Inc. | Method and apparatus for image compression without residue signaling |
| CN105635736B (en) * | 2016-03-14 | 2018-11-13 | 杭州电子科技大学 | A kind of simple and quick merging method |
| CN105898297B (en) * | 2016-04-29 | 2019-03-15 | 上海高智科技发展有限公司 | A kind of fast schema selection method and system based on HEVC |
| CN105791826B (en) * | 2016-05-11 | 2019-03-08 | 南京大学 | A kind of HEVC interframe fast schema selection method based on data mining |
| CN106131546B (en) * | 2016-07-26 | 2018-12-18 | 王靖韬 | A method of determining that HEVC merges and skip coding mode in advance |
| CN108206954B (en) * | 2016-12-16 | 2020-03-06 | 北京金山云网络技术有限公司 | Video coding method and device |
| CN106899850B (en) * | 2017-03-02 | 2020-08-14 | 北方工业大学 | HEVC intra-frame prediction method and device based on SATD |
| CN108347605B (en) * | 2018-01-31 | 2021-09-17 | 南京理工大学 | Quick decision-making method for 3D video depth image quad-tree coding structure division |
| CN108632615A (en) * | 2018-04-09 | 2018-10-09 | 首都师范大学 | A method of the HEVC based on motion vector analysis judges SKIP patterns in advance |
| CN109862354B (en) * | 2019-02-18 | 2023-02-10 | 南京邮电大学 | HEVC (high efficiency video coding) rapid inter-frame depth division method based on residual distribution |
| CN110087087B (en) * | 2019-04-09 | 2023-05-12 | 同济大学 | VVC inter-frame coding unit prediction mode early decision and block division early termination method |
| CN110035285B (en) * | 2019-04-18 | 2023-01-06 | 中南大学 | Depth prediction method based on motion vector sensitivity |
| CN110430427B (en) * | 2019-08-07 | 2021-08-17 | 杭州微帧信息科技有限公司 | SKIP mode coding result-based fast video coding decision method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102595127A (en) * | 2011-01-14 | 2012-07-18 | 索尼公司 | Codeword space reduction for intra chroma mode signaling for hevc |
| CN102595140A (en) * | 2012-03-09 | 2012-07-18 | 北京邮电大学 | Intra-frame prediction video coding method based on image inpainting and vector prediction operators |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2969456A1 (en) * | 2010-12-15 | 2012-06-22 | France Telecom | METHODS AND DEVICES FOR ENCODING AND DECODING AT LEAST ONE IMAGE USING A PREDICT PIXEL SELECTION, CORRESPONDING COMPUTER PROGRAM |
-
2014
- 2014-06-24 CN CN201410288532.XA patent/CN104023233B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102595127A (en) * | 2011-01-14 | 2012-07-18 | 索尼公司 | Codeword space reduction for intra chroma mode signaling for hevc |
| CN102595140A (en) * | 2012-03-09 | 2012-07-18 | 北京邮电大学 | Intra-frame prediction video coding method based on image inpainting and vector prediction operators |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104023233A (en) | 2014-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104023233B (en) | Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) | |
| CN104023234B (en) | Fast inter-frame prediction method applicable to high efficiency video coding (HEVC) | |
| KR101621330B1 (en) | Method and apparatus of transform unit partition with reduced complexity | |
| CN104754357B (en) | Intraframe coding optimization method and device based on convolutional neural networks | |
| EP3553748A1 (en) | Deep learning based image partitioning for video compression | |
| CN104639940B (en) | A kind of quick HEVC method for choosing frame inner forecast mode | |
| TW201811031A (en) | Method and apparatus for template-based intra prediction in image and video coding | |
| CN102763411A (en) | Method and apparatus for encoding video, and method and apparatus for decoding video | |
| US11412220B2 (en) | Texture-based partitioning decisions for video compression | |
| TWI749420B (en) | Method and apparatus of encoding or decoding video blocks with constraints during block partitioning | |
| CN109845256A (en) | Video encoding method/device, image decoding method/device and the recording medium for preserving bit stream | |
| US10992964B2 (en) | Method and apparatus for determining coding tree node split mode and coding device | |
| CN104394409A (en) | Space-domain correlation based rapid HEVC (High Efficiency Video Coding) predication mode selection method | |
| CN101969561A (en) | Intra-frame mode selection method and device and encoder | |
| US20170374361A1 (en) | Method and System Of Controlling A Video Content System | |
| CN103596003B (en) | Interframe predication quick mode selecting method for high-performance video coding | |
| US20200344474A1 (en) | Deep learning based image partitioning for video compression | |
| CN106791828A (en) | High performance video code-transferring method and its transcoder based on machine learning | |
| CN106888379A (en) | It is applied to H.264 arrive the interframe fast video code-transferring method of HEVC | |
| CN102984524B (en) | A kind of video coding-decoding method based on block layer decomposition | |
| Lin et al. | Edge density early termination algorithm for HEVC coding tree block | |
| CN112770120B (en) | 3D video depth map intra-frame rapid coding method based on depth neural network | |
| CN109151467A (en) | Screen content based on image block activity encodes interframe mode quick selecting method | |
| CN110868593B (en) | Video CU fast partitioning based on regional decision tree | |
| CN109889838A (en) | A kind of HEVC fast encoding method based on ROI region |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |